// the find
ericosiu/ai-marketing-skills
Open-source AI marketing skills — growth experiments, sales pipeline, content ops, outbound, SEO, and finance automation
A collection of Python scripts and Claude Code SKILL.md files covering marketing automation: experiment tracking, outbound sales, SEO, content scoring, and financial analysis. Built by a marketing agency to run their own operations, then open-sourced. The target user is a technical marketer or growth engineer who wants to hand these workflows to an AI coding agent rather than build from scratch.
The Growth Engine actually uses real statistics — bootstrap confidence intervals and Mann-Whitney U tests for A/B experiment evaluation, not just 'looks higher to me'. The content scoring pipeline (Expert Panel, Quality Gate) uses recursive expert personas with defined rubrics stored as markdown, which is a practical pattern for getting consistent LLM output. The PII sanitizer with a pre-commit hook is genuinely useful for a repo class where people constantly paste real customer data into scripts. The SKILL.md pattern — drop a file into `.claude/skills/` and the agent picks up the tool context — is a clean way to ship reusable agent tooling that isn't tied to a specific framework.
Each skill module has its own requirements.txt with no shared dependency management, so installing everything across categories is friction and version conflicts are guaranteed. The repo is vendor-coupled in ways the README underplays: RB2B, Instantly, Gong, and Google Search Console are all load-bearing integrations — if you don't have those subscriptions, most of the sales and SEO pipeline is inert. There are no tests anywhere; the scripts appear to be run manually and validated by output inspection, which is a real problem when you're piping LLM output into outbound email sequences. The 'battle-tested on real pipelines generating millions in revenue' claim in the README is unverifiable marketing copy from the agency that built it, and the repo links back to their services page multiple times.